The DB2 monitor keeps DB2SimpleDataSource instances until a configuration
change in the server is detected. However, DB2SimpleDataSource objects are
stateful; in particular they contain the client reroute server list.
Certain configuration issues on the DB2 server side (such as a missing
alternate server name) may cause the client reroute list to become incorrect.
After a failover, the data source then becomes unusable. This implies that
every connection test will fail and the resource will be flagged as unavailable
in RHQ.
In principle, this is desirable because the configuration issue that causes the
problem for RHQ will also cause a problem for the corresponding data sources in
WebSphere. By not resetting the DB2SimpleDataSource automatically, we ensure
that the availability state in RHQ matches the state of the data sources in
WebSphere.
On the WebSphere side, the issue will be solved by restarting the impacted
application servers. However, this will not change the status in RHQ and the
DB2 Monitor resources will continue to be flagged as unavailable.
The obvious solution is to restart the RHQ agent as well, but this may not be
desirable.
Another solution would be to detect the application server restart and to reset
the DB2SimpleDataSource. However, this is not possible because
DB2SimpleDataSource instances are pooled, i.e. a single DB2SimpleDataSource
will be used in the monitoring of multiple application servers (this is to
reduce the number of connections to DB2).
The most effective solution is therefore to add a feature that allows to reset
the state of all DB2 monitors on an agent, so that it is no longer required to
restart the agent.
Original issue reported on code.google.com by andreas.veithen@gmail.com on 22 Nov 2013 at 1:43
Original issue reported on code.google.com by
andreas.veithen@gmail.comon 22 Nov 2013 at 1:43